Larry A. Palmer - US grants

Significant progress in computational neuroscience and neuroengineering requires a multifaceted interdisciplinary research program in several interrelated areas: Mathematical modeling and analysis of neural nets to better understand their collective behavior and capabilities for practical applications. Neurophysiological studies to better understand how retinal information is processed by neuronal assemblies in the striate and extrastriate cortex. Neural vision systems and their VLSI implementation for scene analysis and primitive extraction. Architectures and opto-electronic implementations of self-organizing neural nets partitioned into input/output and internal neurons for supervised and unsupervised learning with stochastic and deterministic state update rules. Higher order processing, in interconnected neural net modules utilizing, sequential and cyclic storage and recall, generalization, for multisensory data fusion and knowledge aggregation. Smart sensing and recognition from sketchy information with emphasis on object recognition including study of object representations that produce distortion invariant recognition. Highly structured associative memory and processing of spoken language. Study of optical materials and devices suitable for realizing artificial plasticity and learning specially in nets with unipolar binary neurons and ternary synaptic weights that facilitate opto-electronic implementations. The present proposal deals with studies to be carried out by a group of faculty with extensive expertise in the above areas, from the schools of Engineering and Medicine. Results of this research are expected to contribute to the development of a new generation of neuromorphic cognitive systems and to outperform more conventional approaches to signal processing. outperform more conventional approaches to signal and

The neurons in areas 17, 18 and 19 of cat are local spatiotemporal operators which extract information about local motion and texture (among other things) from the retinal image. The long range goal of the proposed research is to understand exactly how this is done and to determine the underlying principles, mechanisms and connections which underlie this early visual processing. This proposal concentrates on a portion of the larger problem - namely, the emergence of selectivity for stimulus orientation and motion in simple cells of areas 17 and 18. The work is guided by three related models: (1) simple receptive fields in area 17 act as linear spatial filters of the Gabor form, (2) a new push-pull model of geniculo-simple convergence and (3) Adelson and Bergen's energy model of human motion perception. These ideas will be experimentally tested in areas 17 and 18 of anesthetized cats with the expectation that robust explanations of orientation and direction selectivity will result. It is likely that neural correlates of the early stages of the Adelson-Berger model will be found and that orientation selectivity in simple cells will be attributed to a differential input from populations of LGN cells in a way not previously evident.

DESCRIPTION (Investigator's Abstract): While much is known about the visual properties of neurons in the primary visual cortex, the roles that these cells play as members of cortical circuits responsible for image representation and early processing remain largely unknown. Membership in these circuits necessarily involves interactions spanning several time scales and several levels of organization. The work proposed here is designed to elucidate the circuitry and identify the cellular or synaptic sites of contrast gain control mechanisms and the global suppression of responses elicited by stimuli other than those which excite cells. It is hypothesized that the global suppression is one manifestation of a process by which cellular responses are normalized by total local contrast energy and that this process involves a GABAergic feedback which interacts nonlinearly with feedforward excitation mediated by NMDA receptors for glutamate. This feedback mechanism provides an important key to understanding the representation of complex images by ensembles of neurons. In addition, it is hypothesized that the feedforward connections are self-modifying and provide a variable synaptic gain which accounts for observed pattern-specific adaptation of cortical neurons and shifts of their operating points along the contrast axis. In the course of this work, it is proposed that special emphasis be given to a little-studied class of simple cells located in layer IV whose receptive fields resemble those of cortical afferents most closely. The significance of this work lies in the insights it will provide into global attributes of cortical function. The energy normalization process is of particular importance since a host of recent theoretical studies have emphasized the utility of starting from an energy based representation in order to extract information about motion, texture, stereoscopic depth and shape. At present, the biological foundations of these studies is shaky at best.

IBN: 9634367 PI: Finkel Gain control is a powerful process that is used in many biological and engineering applications. In a videocamera, the photodetectors can only operate over a limited range of light levels, so in order to be able to record in bright daylight and dim internal light, the signal must be re-scaled to stay within a limited range of intensities. A slightly smarter system, used in more expensive devices, is to change the gain (how much the signal is amplified) separately in each local region. Changing the amount of gain in an adaptive manner allows signals to be extracted from ambient noise. The same principles of gain control act in our visual system-- in a simple manner on the pupil to control light entering the eye, but in increasingly more sophisticated ways as one goes to higher levels in the visual cortex. Our goal is to investigate these mechanisms of cortical gain control to better understand how the visual sytem works, and as a means of uncovering new image processing applications. The cortex has the ability to control gain not only based on spatial location, but also in a context-dependent manner. The response to a particular feature, say a short line segment, depends upon what other features are present in the image. If the line segment forms part of a circle or an extended line, the response is increased. Psychophysical studies have identified the conditions under which certain features "pop-out" and become the focus of visual attention. Recent physiological studies have shown that cortical cells are exquisitely sensitive to small changes in the context of the scene. It is believed that response changes are a result of a change in gain of individual cortical cells, but the mechanisms controlling cell gain are poorly understood. We are particularly interested in differential changes of gain among different cells in a neural population. Most features are represented, in the cortex, by a set of graded responses ove r a population of cells. The orientation of a line, for example, is coded by a distribution of responses over cells that prefer various orientations (vertical, horizontal, oblique). We hypothesize that inputs from stimuli in the visual image act to differentially change the gain of cells. A differential change in gain across a population would lead to increased sensitivity--a better estimate of line orientation, or whatever feature is being analyzed. This type of context-dependent gain control could lead to enhanced discrimination of the salient features in an image. We will carry out a series of physiological experiments in cat visual cortex to investigate this hypothesis. We also will conduct a set of computer simulations of various cellular mechanisms to test our model. Finally, we will build a prototype image processing device, based on these findings, and apply it to real images.

DESCRIPTION (Adapted from the Investigator's Abstract): The overarching hypothesis which drives this work is that horizontal connections in the cortex mediate a host of basic visual phenomena and that they do so by modifying the classical receptive fields of cortical neurons in accordance with the visual context surrounding their receptive fields. these processes appear to support segmentation based on differences in disparity, texture, and motion and to support more subtle contextual tasks such as increasing the saliency of such geometry's as smooth curves and illusory contours. The proposed work will identify the regions of space from which contextual effects are manifested, and will test the hypothesis that it is these horizontal cortical connections which are responsible for the contextual effects. The proposed experiments will also discriminate between two hypothetical bases of the contextual effects, i.e., whether it is a change of gain in the classical receptive field mechanism or an actual change in the kernels describing the input-output behavior of the classical receptive field. In comparison with psychophysical effects, the association field of cortical cells will be explored in detail in an effort to account for the special significance of coaxial stimuli outside the classical receptive field and their possible role in the responses seen to illusory contours. In addition, manifestations of the responses elicited from within the classical receptive field in the time domain will be examined. Collectively, these experiments will vastly enhance the understanding of lateral interactions in cortex.

The proposed studies are focused on the temporal integration of excitatory and inhibitory synaptic events in primary visual cortex. The global hypothesis is that cellular and network mechanisms establish windows of time in which synchronous inputs are especially efficacious and asynchronous inputs are suppressed. An additional hypothesis is that these synaptic mechanisms serve to maintain the precision of spike timing of cortical cells at the same level seen in their thalamic afferents. The responses of cells in visual cortex to stimulus pairs spanning a range of temporal separations will be obtained in both extracellular and intracellular experiments. The extracellular studies will establish the full range of these phenomena for cells of all types in all layers. The intracellular studies will be directed at establishing the rules for synaptic integration and their mechanistic substrates at both biophysical and synaptic levels. This will be achieved in part by coupling presentation of visual stimuli with current injection for the identification of synaptic reversal potentials and the time courses of excitatory and inhibitory conductances. These basic experiments will be extended by the use of QX314 to block active membrane channels and altered CI" concentration in the pipettes to shift the reversal potential of inhibition. These studies will be conducted with both simple and complex cells and with cells possessing a wide variety of intrinsic electrophysiological properties. The rules and mechanisms so identified will be related to the reliability and precision of spike trains elicited by temporally rich stimuli. While much has been learned about what computations various areas of cerebral cortex are involved in, our understanding about how these computations are effected is rudimentary at best. The proposed studies will break new ground in term of revealing and understanding mechanisms by which visual cortex processes rapidly changing messages from the two eyes. This will be relevant to studies of cerebral cortex in general and specifically to how visual cortex processes signals elicited by microsaccadic eye movements known to be required for normal sight.

DESCRIPTION (provided by applicant): We propose to continue a broad, interdisciplinary Vision Training Program now in its 29th year. The program includes 29 preceptors in 10 departments and spans many areas: phototransduction (biophysics, molecularbiology);retina (circuitry, computation, neurochemistry, developmental genetics);eye (cataract, myopia, tears);central pathways (physiology, computation);higher processes (psychophysics, cognitive neuroscience, computation);retinal diseases and new genetic therapeutics. The purpose of the Vision Training Program is to promote the intellectual development of outstanding graduate students interested in the visual sciences so that they may ultimately become leaders in their chosen disciplines. Predoctoral students are selected for their excellence by a centralized office of Biomedical Graduate Studies and join a particular "graduate group", mainly Neuroscience, Bioengineering, and Psychology. On Penn's compact campus, graduate education is organized across departments and schools in order to foster multidisciplinary training and collaborative research. Students are trained broadly during their first two years and are attracted specifically to vision research by: (1) formal lectures and laboratory training, (2) weekly lunchtime seminars with research talks by a mix of intramural and external speakers, (3) research rotations in three different laboratories (each followed by a public talk), (4) an annual research retreat with student talks and a scholar in residence program. At this point, all students are exposed to modern research methods (patch clamp, molecular biology, computation, etc...). Collectively, the relevant graduate groups admit about 260 students annually, and of these, about 40 rotate through vision labs. Currently, about 35 are doing their theses in vision labs and of these, 27 are eligible for support by the Vision Training Program. In the past 10 years, the Vision Training Program has trained 19 PhD/MD-PhDs. Most have continued training in excellent labs and many have subsequently assumed positions at prominent research institutions. Through our graduates, the impact of our training program has been widespread, adding significantly'to our understanding of vision in health and disease. Plans for the immediate future include an expanded focus on applied research and translational medicine on-going primarily in the Department of Ophthalmology. Based on continued interest in vision and significant growth of our faculty, we request continued support for 6 predoctoral students per year.